Match Allows Multiple Identical as i32 Statements Without Warning

[oysteinmyrmo]

I tried this code:

#[repr(i32)]
pub enum Something {
    Foo = 0,
    Bar = 1,
    Blarg = 2,
}

fn main() {
    let my_int: i32 = 1;

    match my_int {
        1 => {}
        //1 => {} // Warning: no value can reach this
        x if x == Something::Foo as i32 => {}
        x if x == Something::Foo as i32 => {} // No warning
        x if x == Something::Foo as i32 => {} // No warning
        x if x == Something::Blarg as i32 => {}
        _ => {}
    }
}

rustc seems to happily accept multiple identical cases of enums as i32 in match statements. Using i32 directly gives warnings as expected, but not when representing the enum as i32. There are no warnings in nightly nor in rustc 1.0.0, tested on https://rust.godbolt.org/.

I would expect the same warning for the second x if x == Something::Foo as i32 => {} as for the second 1 => {} in the example below.

At the time of writing https://rust.godbolt.org/ does not allow me to see the disassembly for nightly. But for 1.0.0 the same code is generated for each statement with different offsets.

rustc --version --verbose:

nightly

[hanna-kruppe]

The analysis in rustc that produces this warning doesn’t reason about the guard conditions in $pattern if $guard, only the patterns. Clippy has a lint that should catch identical guards now (since #11223 was closed as fixed) but I can only get the warning to trigger by manually replacing Something::Foo as i32 with 0. So it seems like this should be treated as a bug (false negative) in that clippy lint.

[workingjubilee]

@rustbot transfer rust-clippy